Uni-Bayreuth

Sprungmarken

 

Diploma Thesis

Eddy Sampling Methods for the Measurement of Trace Gas Fluxes

Johannes Ruppert (08/2001-01/2002)

Support: Thomas Foken

Eddy Sampling methods derived from the Eddy Covariance (EC) method are commonly used to measure trace gas fluxes in the boundary layer for investigating exchange processes at the surface-atmosphere interface. This study aims at assessing the different Eddy Sampling methods used when fast sensors are not available and at finding strategies for further improvement by comparing their theoretic principles, related assumptions and sources of error. Methods based on relaxed sampling are indirect methods for flux measurements, because they rely on parameterizations. Disjunct sampling is the basis for relatively new direct methods of flux measurements in the boundary layer. Sources of error stemming from the method itself were investigated and quantified by simulating sampling in Relaxed Eddy Accumulation methods and disjunct eddy sampling. Simulation of ideal Relaxed Eddy Accumulation (REA) were found to yield minimum relative flux errors when a wind deadband of was applied, which is in agreement with results of earlier studies (Oncley et al., 1993). Simulation of Hyperbolic Relaxed Eddy Accumulation (HREA) identified periodic large errors from violation of scalar similarity required for the determination of the hyperbolic deadband and the b-factor. Results from simulation of Disjunct Eddy Covariance (DEC) showed no evidence for systematic errors compared to EC fluxes. The random error was quantified and found to depend mainly on the number of samples acquired when comparing buoyancy flux, humidity flux and CO2 flux results. The simulation of accumulation techniques used ideal definitions of the mean vertical wind velocity . Generally, errors in the simulated ideal Relaxed Eddy Accumulation are smaller than in the simulated disjunct sampling methods using disjunct intervals of 1 s, 3 s, 10 s, or 30 s. Adjustments to the simulation procedure were suggested to more realistically include errors from non-ideal estimates of from preceding data in future investigations.

last modified 2003-12-12